Risk factors for early–onset seizures after stroke: A systematicreview and meta‐analysis of 18 observational studies

Abstract Objectives To systematically evaluate the risk factors of early‐onset seizures after stroke, in order to better provide evidence‐based results for early detection, identification, targeted prevention, and treatment of this disease. Methods PubMed, EMBASE, The Cochrane Library, CNKI, and WanFang databases were searched to collect relevant studies on the risk factors of early‐onset seizures after stroke from January 2010 to January 2020. Meta‐analysis of all included studies was performed by using RevMan version 5.3 and Stata version 14.0 software. Results Eighteen case–control studies with a total sample size of 13,289 cases, including 813 cases with early‐onset seizures after stroke, and 12,476 cases with non‐early‐onset seizures after stroke were included. The results of meta‐analysis showed that cortical involvement [Odds Ratio (OR) = 5.00, 95%Confidence Interval (CI) (2.85, 8.74), p < .00001], cerebral infarction with hemorrhagic transformation [OR = 2.77, 95%CI (1.87, 4.11), p < .00001] and intracerebral hemorrhage [OR = 1.83, 95%CI (1.13, 2.97), p = .01]‐related factors showed greater association with the occurrence of early‐onset seizures after stroke. Conclusions These findings suggest that cortical involvement, intracerebral hemorrhage, and cerebral infarction with hemorrhagic transformation are important predictors and risk factors for early seizures after stroke, while the patient's gender, age, NHISS score, alcoholism, smoking, high blood pressure, diabetes, atrial fibrillation, dyslipidemia, receiving surgical treatment, and reperfusion therapy showed no association with the occurrence of early‐onset seizures after stroke.


| INTRODUC TI ON
Early-onset seizures after stroke (ES) is defined as seizures that occur within 7 days after stroke onset (Fisher et al., 2014) and is regarded as a common complication after stroke. Biffi et al. (2016) have pointed out that about 10% of patients included in the study have diagnosed with early seizures within 7 days of intracerebral hemorrhage. Beghi et al. (2011) conducted a large multicenter study and found that the incidence of acute symptomatic seizures was 6.3% in patients with ischemic or hemorrhagic stroke. Van Tuijl JH and other researchers (Tuijl et al., 2018) have revealed that the disability and mortality rates in ES patients were significantly higher than those of non-epileptic patients. The occurrence or recurrence of seizures symptoms in stroke patients may lead to unfavorable functional prognosis (Bentes et al., 2017;Huang et al., 2014), poor quality of life (Zelano, 2020), and higher mortality rate (Mohamed & Kissani, 2015). This negative effect not only exists in elderly stroke patients but also in young patients (Arntz et al., 2015).
At present, there are many studies (Abraira et al., 2020;Bian, 2017;Shehta et al., 2018) that discussed the risk factors for ES after stroke. The risk factors for individuals with ES included baseline characteristic factors, lifestyle-related factors, underlying diseases-related factors, brain injury-related factors, and whether to receive surgery. For example, Abraira et al. (2020) have considered cerebral hemorrhage and cortical hemorrhage as risk factors for ES. Shehta et al. (2018) have believed that intracerebral hemorrhage, cortical lesions, and large lesion size as risk factors for ES. And Bian (2017) has believed that infarct lesions of 2-5 cm diameter and cortical lesions are related to the occurrence of ES.
However, due to heterogeneity and diversity of the disease itself, the results of each study are different and occasionally contradictory. For example, Szaflarski et al. (2008) found that younger patients have a higher incidence of ES. However, researchers such as Procaccianti et al. (2012) and Wang et al. (2013) hold different views and believe that age is not a risk factor for ES. In terms of stroke severity factors, although Mohamed and Kissani (2015) and Goswami et al. (2012) found that the severity of stroke can be used as a risk factor for ES, especially Procaccianti et al. (2012) believes that ES may be considered a marker of stroke severity; Wang et al. (2013) believes that stroke patients are not associated with ES symptoms. The factors related to stroke subtypes are more controversial. Some studies (Abraira et al., 2020;Bladin et al., 2000;Goswami et al., 2012) have reported that hemorrhagic stroke is an independent risk factor for ES, Hundozi et al. (2016) has found that hemorrhagic and ischemic stroke patients have the same incidence of ES after stroke, while Aiwansoba and Chukwuyem (2014) has found that cerebral infarction is more related to ES. Therefore, this meta-analysis was conducted on the risk factors of ES that are controversial and systematically evaluates the main risk factors that occur in order to provide better decision recommendations for guiding clinicians in early identification, prevention, diagnosis, and treatment.

| Literature retrieval strategy
The current meta-analysis was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) (Reference). PubMed, EMBASE, the Cochrane Library, CNKI, and WANFANG databases were searched to identify relevant studies. The search was conducted from January 2010 to January 2020 using the search terms "after stroke", "cerebral hemorrhage", "brain ischemia", "early seizures", "risk factors", and "factors".
The search words and keywords were also joined for conducting the search. The reference lists of eligible publications were manually checked to identify any other potential studies. 3. Diagnostic criteria: International League Against Epilepsy for ES is defined as the appearance of seizures symptoms within 7 days after a stroke, and the criteria for appearance of seizures symptoms within 14 days (2 weeks) after a stroke are also adopted by the researchers (Denier et al., 2010;Hundozi et al., 2016;Menon & Shorvon, 2009;Wang et al., 2013). Therefore, the ES diagnostic criteria used in this study are seizures within 14 days (2 weeks) after stroke, considering that as much sample size as possible should be taken.

| Document extraction
The literature was screened according to the inclusion and exclusion criteria, and the following information was extracted: the first author's name, publication time, nationality of study population, total number of samples in case and control groups, average age, possible risk factors, details of exposure factors, etc.

| Study quality of risk of bias
Two investigators (MA Sitian and WANG Huan) have independently evaluated the risk of bias of included studies. If there are inconsistent opinions, then a third author (YANG Yongfeng) was contacted for decision. Third-party opinions will be looked out if still differences exist. The bias risk evaluation of case-control studies was done using the NOS scale in order to evaluate the quality of literature included, in which a score of 9 points, ≥ 7 points were considered as high-quality, 4 to 6 points were considered as medium-quality articles, and ≤3 points were considered as low-quality articles ( Table 2).

| Statistical methods
Statistical analysis was performed using Review Manager version 5.3 software. I 2 was used to evaluate heterogeneity. When p > .1 and I 2 < 50%, a fixed-effects model is used. Otherwise, a randomeffects model was used. A subgroup analysis was performed based on factors such as sample size, country, and region to find the source of heterogeneity. In case-control studies, the odds ratios (OR) was used as the effect scale and its 95% confidence intervals (CI) was calculated at the same time, with p ≤ .05 as statistically significant difference. The man differences (MD) were used as the effect scale for continuous variables, and its 95% CI was calculated at the same time. The difference was statistically significant with p ≤ .05. Sensitivity analysis was used to calculate combined OR value and 95% CI, and compared the two sets of results to show whether the results are stable. Egger funnel chart was drawn using Stata version 14.0 software, and publication bias was evaluated by Egger p value result and whether the funnel chart was symmetrical or not. If p > .10 and funnel chart showed no obvious asymmetry, this indicated no obvious publication bias. Otherwise, publication bias was indicated.

| Factors related to baseline characteristic
A meta-analysis of baseline characteristic factors such as the patient's gender, age, NHISS score, and ES occurrence (Table 3) has been carried out. The results show that there is no heterogeneity in the characteristic factors of gender and age (I 2 = 0, p > .10), so the fixed effects model was adopted; the characteristic factors of NIHSS at admission were heterogeneous (I 2 = 92%, p < .000 01), so the random effects model was adopted. Meta-analysis results show that (Table 3): the above-mentioned baseline characteristic factors are not statistically significant in association with ES, suggesting that gender, age, and NIHSS at admission are not risk factors for ES.
According to the difference of the sample size, a subgroup analysis about the characteristic factor of NIHSS at admission has been carried out ( Figure 2) and the results showed that: when the sample size is >500, the analysis result is statistically significant [MD (95%Cl) = 4.58 (2.62, 6.53), p < .000 01]. Therefore, it can be determined that when the sample size is greater than 500, the degree of neurological deficit is correlated with the occurrence of ES.

| Factors related to lifestyle habits
Four studies (Abraira et al., 2020;Bian, 2017;Goswami et al., 2012;Zeng, 2018) evaluated the relationship between alcoholism and ES (Table 3). A total of 2,933 samples were included, of which there were 161 cases in the ES group after stroke, and 2,772 cases in the no-ES group after stroke. After extracting the data and calculating it, I 2 = 92%, p < .000 01, suggested the existence of heterogeneity.
Research suggests that there is significant heterogeneity in alcoholism factors, so subgroup analysis is carried out according to sample size and different countries (Table 4 and Table 5), in which the results showed no changes and had no effect on the outcome indicators.

| Related factors of basic diseases
Meta-analysis on the relationship between the underlying disease factors related to hypertension, diabetes, atrial fibrillation, and dyslipidemia and the occurrence of ES (Table 3) was conducted.
The results showed that there were no heterogeneity in the two factors of hypertension and dyslipidemia (I 2 = 0, p > .10), and the fixed-effects model was adopted. While the two factors of diabetes and atrial fibrillation are heterogeneous (I 2 = 55%, p < .10), and the F I G U R E 1 Literature search process and results. * The searched databases and detected documents are as follows: PubMed (n = 445), EMbase (n = 217), The Cochrane Library (n = 19), CNKI (n = 232), WanFang data (n = 345) random-effects model was adopted. The results of meta-analysis showed no statistically significant association between the above factors and ES, suggesting hypertension, diabetes, atrial fibrillation, and dyslipidemia as not risk factors for ES.
The study found significant heterogeneity in the factors of diabetes and atrial fibrillation, so we further conducted a subgroup analysis of different sample sizes and countries (Table 4 and Table 5).
This can partially explain the source of heterogeneity related to diabetes, but the results of atrial fibrillation still remain unchanged.
The results showed that the composition ratio of cortical damage in ES group was greater than that in the no-ES group [OR (95% Cl) = 5.00 (2.85, 8.74)]. The difference was statistically significant (p < .01), indicating cortical injury as a risk factor for ES after stroke.
The study found that there is significant heterogeneity in cortical damage factors, so we further conducted subgroup analysis according to different sample sizes ( Figure 4) and regions (Asia/Europe/ Africa). The results showed that the sample size can partially explain the source of heterogeneity of cortical damage factors. However, the results of subgroup analysis in different regions were unchanged.

Hemorrhagic transformation
Six (Abraira et al., 2020;Gao, 2016;Mohamed & Kissani, 2015;Pezzini et al., 2013;Procaccianti et al., 2012;Shehta et al., 2018) studies reported hemorrhagic transformation and occurrence of ES ( Figure 5). The data suggested that no heterogeneity (I 2 = 0), and fixed-effects model was used. The results showed that OR (95% Cl) = 2.77 (1.87, 4.11), suggesting statistical significance and indicating that hemorrhagic transformation is a risk factor for ES.  Note: In the "Selection" and "Exposure" categories, a quality item of a study can be rated at most one "★", and for the "Comparability" category, at most two "★".

| Factors related to treatment
Two studies (Bian, 2017;Yang, 2012) have investigated the relationship between receiving surgical treatment and occurrence of ES ( suggesting surgical treatment as not risk factor for ES.
Two articles (Abraira et al., 2020;Hundozi et al., 2016) carried out a study on the relationship between patients receiving reperfusion therapy and the occurrence of ES (

| Sensitivity analysis
For each risk factor, fixed-effects and random-effects models were used to calculate the combined OR value and 95% CI ( Table 7). The results showed that the OR (95% CI) value of fixed-effects model of alcoholism was 1.83 (1.30-2.58), which showed statistical significance. While the OR (95%CI) value of random-effects model was 2.46 (0.46-13.14), which included invalid value 1 and showed no statistical significance. This suggested that the research results of this factor of alcoholism are unstable. The fixed-effects and random-effects results of the remaining risk factors are close, indicating that the conclusion of the study was relatively steady.

| Publication bias analysis
The risk factors that included in more than five articles were selected and used Egger method to conduct publication bias test and statistical results of Egger test p-value (Table 8). The three risk factors of cortical injury, cerebral infarction with hemorrhagic transformation were taken, and stroke subtype as example to draw Egger funnel diagram (Figures 8-10). The Egger test results in Figures 8,   9, and 10 showed that their p-values are equal to 0.300, 0.942, and 0.502, respectively, and no obvious asymmetry in the funnel chart was observed, indicating no obvious publication bias.

| D ISCUSS I ON
The pathogenesis of ES after stroke still remained unclear (Zelano, 2020). The risk factors for this symptom have not been fully confirmed. Therefore, what is of great significance for the early iden- Another study found (Kamp et al., 2012) that damage to the stroke cortex in rodent models caused changes in ion channels within a day.  Some studies (Alberti et al., 2008;Bian, 2017;Mohamed & Kissani, 2015) believe that cerebral infarction with hemorrhagic transformation is also the cause of ES. The pathogenesis of ES may be related to the persistent presence of a large amount of glutamate and the release of high levels of excitotoxicity neurotransmitters in the ischemic injury area (Rodríguez Lucci et al., 2018). Procaccianti et al. (2012) believe that cerebral ischemia with hemorrhagic transformation can cause the occurrence of ES. They found that patients with ischemic stroke may be due to the effect of blood degradation products on cortical neurons, which leads to the occurrence of ES. The study included six literatures review their relevance, results showed that with hemorrhagic transformation of cerebral infarction is a risk factor of ES.  Goswami et al. (2012) and Lekoubou et al. (2020) might explain the increased incidence of ES in patients with cerebral hemorrhage.
An earlier study (Berger et al., 1988) suggested that almost all intracerebral hemorrhage associated seizures occurred within a short time after the onset of the intracerebral hemorrhage. Some studies have shown that about 50%-70% of ES will occur in the first 24 hr (Gilmore et al., 2010;Vespa et al., 2003). This meta-analysis included 11 studies on stroke subtypes (cerebral hemorrhage/cerebral infarction) and the occurrence of ES. The study found cerebral hemorrhage as a risk factor for ES.
Regarding baseline characteristic factors such as age and NIHSS  (Beghi et al., 2011) have suggested that hyperlipidemia is a protective factor for hemorrhagic stroke. Studies (Chen et al., 2016;Goswami et al., 2012;Shmuely et al., 2017) have shown that there is a high correlation between hypertension, diabetes, heart disease and the occurrence of ES. Nass et al. (2019) believe that epileptic seizures can cause abnormal changes in blood pressure, which may be caused by epileptic activity that stimulates or inhibits the function of the central nervous system and spreads to different neuronal networks. Therefore, the inconsistency of the conclusions requires further multi-center and large-sample research. Regarding reperfusion therapy factor, we conducted a meta-analysis and found that it has no correlation with the occurrence of ES. In this study, only two articles that meet the requirements were included. Therefore, more prospective studies are needed to confirm whether reperfusion therapy is a risk factor for ES for early identification or prediction.
With regard to alcoholism, due to inconsistent results of fixedeffects and random-effects calculations, the results of metaanalysis remained unstable. This is because too few study samples were included, or the definition of alcoholism is still inconsistent or F I G U R E 9 Egger funnel chart (hemorrhagic transformation) F I G U R E 1 0 Egger funnel chart (stroke subtype) not described. Researcher Zhang et al. (2014) have found alcoholism (OR = 1.70, 95% CI = 1.23-2.34, p < .01) as a risk factor for ES after stroke through systematic review and meta-analysis. But our study showed no statistical significance between alcoholism and ES.
Therefore, whether alcoholism is related to the occurrence of ES, whether it acts as a risk factor for ES and the underlying mechanism of alcoholism that leads to seizures requires further research.

| Limitations
There are still many deficiencies that should be acknowledged in this study and are as follows: (1)

| CON CLUS ION
In summary, the results of this meta-analysis indicate that cortical injury, cerebral infarction with hemorrhagic transformation, and cerebral hemorrhage are closely related to the occurrence of ES, and are risk factors and important predictors for the occurrence of ES.
Medical staff can refer to the results of this study to identify early, targeted risk groups of early-onset seizures after stroke, reduce early incidence of seizures in patients after stroke and thus improve the quality of life of patients and their families.

ACK N OWLED G EM ENTS
Not applicable.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests. Xiaoxuan helped perform the analysis with constructive discussions.

E TH I C A L A PPROVA L
Not applicable.

CO N S E NT FO R PU B LI C ATI O N
Not applicable.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.2142.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analyzed during this study are included in this published article and its supplementary information files.